Is it an AT90S8535 or something? I would use ATmega16 instead. They both have ADC's anyway. It's simply the matter of setting a few registers and reading some other registers later on.
I don't follow you quite well either. How exactly did you statically "define" the brightness of the variable light source in past tests? I'm curious about how you interpret the brightness into anything.
The picture in the attachment is what I would do to find the brightest spot. A four quadrant matrix. panels absorb light and helps determining direction of stronger source. Compare average values in two directions to construct a two-dimentional ambient brightness map.
i.e,
if (A+B)/2 > (C+D)/2 Then brightest spot is to the front
if (A+C)/2 > (B+D)/2 Then brightest spot is also to the left
Whether or not you pass some parameters of any kind from former tests, I don't think these "defined" values are going to be of assistance. Coz the problem with absolute values is that they make things less adaptive. Once the environment changes, even slightly, they will have no idea what to do.
Actually you shouldn't have needed to run any "past test", just compare different values and use the relationship, instead of looking into the exact numbers trying to figure out something that won't be the same for ever.